Rewrite initialization of builtins to use the BuiltIn trait

[jedel1043]

We currently aren't using the BuiltIn trait to check the initialization of builtins, meaning any struct that implements a init function that returns a (&str, Value, Attribute) tuple can be used to initialize a global builtin.

This PR changes the following:

• Refactors BuiltIn::init to return only a JsValue and converts BuiltIn::attribute into a constant.
• Refactors the initialization of all builtins to check for an implementation of the BuiltIn trait.
• Creates a globals! macro to avoid the boilerplate.
• Refactors all builtins to conform to the new trait.

[github-actions]

Test262 conformance changes:

Test result	master count	PR count	difference
Total	80,930	80,930	0
Passed	33,192	33,192	0
Ignored	15,898	15,898	0
Failed	31,840	31,840	0
Panics	0	0	0
Conformance	41.01%	41.01%	0.00%

[raskad]

Seems like a good idea to me.

Not sure if we need this macro, as I personally think macros always negatively impact readability and probably(?) also compile times.

[jedel1043]

Not sure if we need this macro, as I personally think macros always negatively impact readability and probably(?) also compile times.

I think this is a good case of use for macros. The alternative would be to manually write 26 calls to init_builtin in succession, which a macro is perfect for, and it clearly defines what are the global constructors of our context, making the implementation of new builtins as easy as putting them in the globals! list and solving any compile errors.